The post-2020 redistricting cycle was likely the last one in which political mapmakers drew electoral maps manually, one district at a time. After the 2030 Census, mapmakers may rely instead on the emerging science of computational redistricting, using algorithms to generate and evaluate statewide maps that attempt to optimize compliance with multiple redistricting criteria simultaneously. This new approach permits mapmakers to sift through billions of high-performing alternatives in search of the best available options—far outstripping what any previous mapmaker could do by hand.
This Article aims to identify the strengths and limits of computational redistricting. It does so with an eye to both theory and practice. The Article first outlines the basic theoretical parameters of a computational approach. Then, drawing on the coauthors’ experience as counsel and consulting experts in the recent litigation challenging Wisconsin’s state-legislative maps, the Article describes how a particular version of a computational approach was used to address Wisconsin’s extreme partisan gerrymandering. In so doing, the Article offers practical guidance to future litigators and mapmakers who will face similar challenges across the United States in the post-2030 redistricting cycle.
The Article concludes that computational redistricting has significant advantages over prior approaches because it meaningfully advances mapmakers’ ability to balance the various redistricting criteria set forth in federal and state law. Moreover, the improved quality of maps that a court may adopt to correct a constitutional violation may drive political actors to reach their own settlements—as occurred in Wisconsin. As the Wisconsin litigation also reveals, however, computational redistricting cannot produce a singularly “neutral” or “apolitical” map with the push of a button. Computational redistricting still requires numerous judgment calls about how best to construe, measure, and weight the various criteria. More fundamentally, redistricting implicates normative considerations bearing on the relationship between the electors and the elected in a republican form of government. Those issues are not “solvable” by applying increased computational power.
